The Journal of Wildlife Management 79(2):180–194; 2015; DOI: 10.1002/jwmg.826
Research Article
Influence of Landscape Features on Spatial
Genetic Structure of White-Tailed Deer in
Human-Altered Landscapes
ALEXANDRA LOCHER,1 Biology Department, Grand Valley State University, 1 Campus Drive, Allendale, MI 49401, USA
KIM T. SCRIBNER, Department of Fisheries and Wildlife, Michigan State University, Room 13 Natural Resources, East Lansing, MI 48824, USA
JENNIFER A. MOORE, Biology Department, Grand Valley State University, 1 Campus Drive, Allendale, MI 49401, USA
BRITTANY MURPHY, Department of Fisheries and Wildlife, Michigan State University, Room 13 Natural Resources, East Lansing, MI 48824,
USA
JEANNETTE KANEFSKY, Department of Fisheries and Wildlife, Michigan State University, Room 13 Natural Resources, East Lansing, MI
48824, USA
ABSTRACT Predictive relationships between estimates of functional population connectivity and physical
and biotic landscape features can provide important insights into present and future population responses to
human-mediated landscape change. Quantification of associations between landscape features and dispersal
or genetic surrogates such as gene flow among areas can be particularly challenging for continuously
distributed and highly mobile wildlife species. We assessed the relative influence of natural and humanaltered landscape features on white-tailed deer (Odocoileus virginianus) spatial genetic structure (SGS) in
southern Michigan (USA) using 7 microsatellite markers assayed for 326 adult individuals from 21
contiguous counties (33,284 km2). We used previously collected telemetry data to quantify probabilities of
habitat occupancy and seasonal movements that allowed selection and weighting of landscape features to
create habitat suitability indices (HSI). We assigned individuals to groups (n ¼ 13) for statistical analyses
quantifying relationships between measures of SGS (response variable) with Euclidean distance, least cost
distances parameterized using HSI, and presence of natural (rivers) and man-made (roads) barriers to
dispersal. Over the entire study area, genetic differentiation was significant (mean Fst ¼ 0.019, P < 0.001) and
increased with increasing inter-group geographic distance (r2 ¼ 0.381; P < 0.05). We identified features in
the landscape matrix between groups including rivers, high traffic roads, and habitats of intermediate HSI as
inhibiting gene flow. Low HSI was associated with low between-group Fst and appeared to facilitate gene
flow. Quantification of the relative importance of man-made barriers (roads) and habitat suitability to SGS
for white-tailed deer emphasizes the importance of joint use of ecological and genetic analyses in conservation
and control efforts for abundant and mobile wildlife species. Ó 2014 The Wildlife Society.
KEY WORDS functional connectivity, gene flow, habitat suitability, landscape genetics, least cost distance, whitetailed deer.
Landscape composition and configuration typically vary over
space and time (Wagner and Fortin 2005) because of natural
processes that have occurred over historical and contemporary
time scales (Anderson et al. 2010) and because of recent human
activities (D-Eon et al. 2002). Increasingly, landscape changes are
occurring on ecological time scalesbecause human activities have
left a more recent and expansive footprint on landscape features
than have natural processes (Kareiva et al. 2007). These recent
events including creation of roads (Forman and Alexander 1998,
Balkenhol and Waits 2009) or changes in environmental
conditions and land-use (Geffen et al. 2004, Coulon et al.
2006) affect probabilities of habitat occupancy and rates of gene
Received: 11 November 2013; Accepted: 5 November 2014
Published: 26 December 2014
1
E-mail: lochera@gvsu.edu
180
flow among wildlife populations. Understanding how landscape
features and aspects of a species’ ecology affect movement patterns
and population connectivity is a major challenge facing wildlife
conservation and management programs attempting to maintain
ecological integrity (Andreasen et al. 2001), functional connectivity (Beier et al. 2008), population viability (Hanski 2002), or to
minimize human conflicts (Henderson et al. 2000). Further,
landscape features can act as agents of selection (Manel et al.
2010). Therefore, characterizing rates of dispersal through
heterogeneous landscapes is important to detect selection along
environmental gradients (De Mita et al. 2013), and is increasingly
possible because genomics data is becoming available for many
organisms, including ungulates (Haynes and Latch 2012).
Natural and anthropogenic processes have affected physical
and biotic landscape features that concurrently have
influenced wildlife population demography and individual
movements (Clobert et al. 2009). These processes can, in
The Journal of Wildlife Management
79(2)
turn, influence levels of genetic variability within populations
and the degree of spatial variation in population gene
frequency (i.e., operational definition of spatial genetic
structure [SGS]; Waples and Gaggiotti 2006, Boulet et al.
2007). For instance, features of the physical and biotic
environment such as habitat quality may affect the
probability of movements such as dispersal (Bowler and
Benton 2005). Habitat quality considers multiple components collectively (i.e., food, cover) to evaluate the capacity of
an area to support individuals. As habitat quality varies
spatially, individuals are likely to remain in areas (patches) of
high habitat quality (Knowlton and Graham 2010) and
disperse into or through areas of lower habitat quality. Thus,
if individuals are philopatric to natal areas, expectations are
that areas of comparatively higher and lower habitat quality
will be characterized by higher and lower measures of genetic
diversity, respectively (Epperson 1993), including mean
relatedness. For example, areas of higher habitat quality may
support higher population densities and individual fitness.
Likewise, areas of lower habitat quality may demonstrate
lower levels of genetic diversity due to smaller population
sizes and decreased fitness (Morris et al. 2004). These lower
quality areas may also exhibit higher levels of gene flow as
individuals disperse through and potentially seek areas of
higher quality habitat (Morris et al. 2004). Other landscape
features such as roads (Balkenhol and Waits 2009) and rivers
(Blanchong et al. 2008) may present barriers to dispersal, and
thus affect SGS.
For species that disperse long distances or are widely
distributed throughout the landscape, measures of genetic
distance or spatial variance in allele frequency (e.g.,
F-statistics; Weir and Cockerham 1984) are often used to
characterize gene flow among groups of animals captured
from different locales (Cushman et al. 2006, McRae and
Beier 2007, Purrenhage et al. 2009). Accordingly, landscape
genetic studies (Manel et al. 2003; Storfer et al. 2007,
Holderegger and Wagner 2008) using estimates of SGS and
remotely sensed spatial data describing landscape features are
increasingly emphasized as an effective means to quantify the
relative importance of factors affecting connectivity among
populations (Cushman et al. 2006, Hall and Bessinger 2014).
As wildlife populations are increasingly affected by human
activities and landscape changes (Fahrig 2007), understanding the complex interactions between measures of SGS,
geographic distance, and landscape permeability is critical
for management or conservation planning (Minor and
Urban 2007, Kool et al. 2010), and population control
(Côté et al. 2004).
The genetic signatures of population responses to human
activities are often dependent on a species’ vagility
(Landguth et al. 2010a,b). White-tailed deer (Odocoileus
virginianus) are highly mobile habitat generalists
(Hirth 1977)that have successfully adapted to humanaltered landscapes including areas interspersed with forest
cover, edge (Marchinton and Hirth 1984, Shi et al. 2006),
agricultural crops that supplement natural forage (Gladfelter 1984), and urban areas (Blanchong et al. 2013). Whitetailed deer population abundance has increased from
Locher et al.
Deer Spatial Genetic Structure
historical (pre-settlement) levels in southern Michigan,
USA(Michigan Department of Natural Resources
[MDNR] 2009). In southern Michigan and many areas
in the north-central United States and southern Canada,
historically forested landscapes have been fragmented by
agricultural practices and human developments such as
transportation networks that have created an increasing
footprint of urbanization over the past 150 years (Hawbacker et al. 2006). High levels of abundance do not alone
imply that dispersal across anthropogenically altered landscapes is extensive. Therefore, management and control
activities for natural populations of ecologically and
economically important species such as white-tailed deer
necessitate a greater understanding of how landscape
features affect movements (With et al. 1997, Crooks and
Sanjayan 2006).
Our general objective was to quantify relationships between
natural landscape features and features associated with past
and present human activities and SGS of white-tailed deer in
southern Michigan. We hypothesized that landscape
features reflecting the quality and spatial heterogeneity
(Diefenbach et al. 2008) and degree of habitat fragmentation
(Long et al. 2005), that are common in the north-central
United States and Michigan are associated with SGS.
Specifically, our working hypotheses were 1) habitat
suitability within patches occupied by white-tailed deer
was positively related to genetic diversity and SGS, 2)
measures of habitat suitability and landscape resistance
between populations were better predictors of genetic
diversity and SGS than geographic distance alone, and 3)
natural landscape features including rivers as well as
prominent and recent human construction (highways)
were barriers to movements such as dispersal, as reflected
by estimatesof SGS.
STUDY AREA
We conducted the study across 21 counties (33,284 km2) in
southern Michigan. The area is a fragmented mosaic of landcover and land-use practices, dominated by agricultural
cropland planted predominately in corn and soybeans,
particularly in the central region of the study area. Natural
vegetation including deciduous upland and lowland hardwood habitat types were more abundant in the western,
eastern, and northern portions of the study area. The
growing season (i.e., the average annual accumulation of
daily mean temperatures >5.68 C) was 150 days, generally
occurring between mid-May to mid-October (Sommers
1977). Average annual snowfall was 120.2 cm between 1999
and 2000 (Midwestern Regional Climate Center,
Champaign, Illinois, USA). Deer densities within the study
area range from 15 to 40 deer/km2; densities were higher in
areas with limited hunting access (MDNR 2010). Because
winter thermal cover may not be necessary for deer in this
area (Torgersen and Porath 1984), individuals tend to be
non-migratory between summer and winter ranges
(Pusateri 2003, Hiller 2007).
181
METHODS
Development of Indices of Habitat Suitability
Habitat suitability indices are frequently used to identify
characteristics of areas that are selected by wildlife (Morrison
et al. 2006). Radio-telemetry data are often used to develop
habitat suitability indices (Aebischer et al.1993, Epps et al.
2007, Chietkiewicz and Boyce 2009, review of applications
with genetic data in Spear et al. 2010). We used previous
radio-telemetry data on white-tailed deer within 2
portions of our study area (Pusateri 2003, Hiller 2007) to
identify and weight landscape features that have been
identified as features important for determining deer habitat
selection during all seasons in southern Michigan (Miranda
and Porter 2003, Hiller 2007). The studies used 59 adult deer
captured during winter 2001–2002 (Pusateri 2003, Michigan
State University All-University Committee on Animal Use
and Care approval 01/01–001-00) and 66 adult deer captured
during winter 2004–2006 (Hiller 2007, Michigan State
University All-University Committee on Animal Use and
Care approval 01/04–006-00) located 2–5 times/week
during the entire year using triangulation methods (White
and Garrott 1990). The number of location estimates
averaged 132/deer (Hiller 2007).
We obtained characterization of white-tailed deer habitat
suitability in southern Michigan landscapes using land cover
classifications from the Integrated Forest Monitoring,
Assessment, and Prescription (IFMAP) data (MDNR
2003) at a resolution of 30-m 30-m in ArcGIS 9.3.1
(Environmental Systems Research Institute, Redlands,
California, USA). We used the specified resolution because
the IFMAP data was classified from Landsat Thematic
Mapper satellite imagery produced in 30-m 30-m cells.
We estimated a composite habitat suitability value based on
the proportion of area classified as urban, coniferous, upland
deciduous, woody wetland, and agricultural edge within a
circular moving window with a radius of 900 m. A 900-m
radius window was comparable to the home range size of a
typical adult female deer in Michigan (Pusateri 2003, Hiller
2007). The value assigned to each pixel was the habitat
suitability within a hypothetical deer’s home range if it were
centered on any 30-m 30-m pixel within the study area.
Habitat suitability was quantified as:
HSI ¼
ð4CONIFÞ þ ð2DECI DÞ þ ð2AG EDGEÞ þ U RBAN þ W ET L
10
where the habitat suitability index (HSI) was the weighted
average of CONIF, DECID, AGEDGE, URBAN, and
WETL. Weighting coefficients (Spear et al. 2010) were
assigned based on a ranking of importance (relative
occupancy time within each habitat type) to deer habitat
selection in southern Michigan. The variable CONIF was
the suitability index on a scale of 0–100 of the proportion of
coniferous forest type within the evaluation area. The
minimum threshold for high suitability was 10% of the home
range comprised of coniferous forest types (Pusateri 2003,
Hiller 2007). The variable DECID was the suitability index
for the proportion of deciduous forest cover; optimal
182
conditions occurred between 20% and 70% (Hiller 2007).
AGEDGE was the suitability index quantifying the
proportion of agricultural row crops within 180 m from a
forest edge. Agricultural crops provide an abundant food
source for deer (Gladfelter 1984), and comprise the major
portion of their diet (Nixon et al. 1970) and home range
(Pusateri 2003, Hiller 2007) in the mid-west agricultural
region of the United States. Thus, suitability of AGEDGE
increased linearly with percent of agricultural edge. Deer
generally do not select residential or urban areas (i.e., areas
with man-made structures, roads, golf courses, residential
parks; Pusateri 2003, Hiller 2007). However, some lowdensity residential areas may provide food sources for deer
(Miranda and Porter 2003). Therefore, the suitability of
URBAN was based on a negative relationship to the percent
of residential/urban area within the moving window. The
variable WETL was an index quantifying the ideal woody
wetland composition within the window, suitability was
optimal between 10% and 25% (Pusateri 2003).
We also included in analyses presence–absence of
natural landscape features including rivers (Blanchong
et al. 2008) and roads (Balkenhol and Waits 2009,
Corlatti et al. 2009) that have been found to affect
SGS. We classified highways according to levels of
use (traffic volume) based on information acquired
from the Michigan Department of Transportation
Traffic Monitoring Information System (http://www.
michigan.gov/mdot-tmis). Specifically, we classified
major (4 lane) highways as high-traffic (>50,000
vehicles/24 hr), medium traffic (20,000–50,000 vehicles/
24 hr), or low traffic(<20,000 vehicles/24 hr). We also
used presence of all water bodies classified as rivers
according to Michigan hydrology maps (Michigan Center
for Geographic Information 2009).
Sampling for Genetic Analysis
Several aspects of study design and genetic analysis can
improve inferences regarding the influences of landscape
features on spatial genetic structure (Anderson et al. 2010),
including the spatial extent, lag, and grain over which
observations are collected and analyses conducted (Dungan
et al. 2002, Cushman and Landguth 2010). Locations of
individual adult (1.5 yr) deer samples (n ¼ 326; 44% female
and 56% male) collected over the period 1998–2000 were
spatially referenced based on the location of harvest
(township, range, section) provided by hunters.
We assigned individuals to groups that were spatially
arrayed over contiguous geographic areas to achieve
replication for statistical comparisons between measures of
spatial genetic structure, inter-group Euclidean distance,
inter-group measures of habitat quality (HSI), and
presence–absence of dispersal barriers (roads and rivers).
Specifically, we assigned individuals to 13 contiguous groups
that we defined a priori rather than a posteriori (e.g., using
clustering algorithms such as BAPS; Corander Marttinen
2006) because deer in Michigan (e.g., Blanchong et al. 2007)
and elsewhere in the Midwestern United States (e.g.,
Robinson et al. 2012) are continuously distributed and
The Journal of Wildlife Management
79(2)
exhibit a pattern of isolation-by-distance (Wright 1943).
We used groups rather than individuals for several reasons.
The first is precedence in the landscape genetics literature
(Koyghobadi et al. 2005, Goldberg and Waits 2010, Murphy
et al. 2010, review in Storfer et al. 2007). Secondly, our deer
samples were spatially referenced only to a single (harvest)
location, which makes assigning accurate weights, such as
habitat suitability difficult. Also, uncertainty may exist
regarding the reported harvest locations of individuals, and
our harvest locations were reported only to the level of a
section (a 2.6-km2 area). These problems are less of a concern
when characterizing habitat patches and habitat suitability
for a group of individuals inhabiting those patches. Mean
area sampled per group was 734 km2, range 259–1,277 km2;
Fig. 1). We assumed the groups were a representative sample
to understand the spatial genetic structure of deer within the
study area.
We identified boundaries of deer groups in ArcMap using
the minimum convex polygon (MCP) method with Hawth’s
Analysis Tools (Beyer 2004) and connecting points of the
outer coordinates of clustered (i.e., based on proximity) deer
samples. Group size averaged 25 individuals (range 14–42).
Mean inter-group distance was 101 km (range 30–189 km).
We calculated the average HSI within the polygon defining
the boundary of each group to quantify mean habitat
suitability associated with landscapes within each group
polygon.
Genetic Analysis
We extracted DNA from archived tissue samples collected
from white-tailed deer using QIAGEN DNeasy extraction
kits (Qiagen, Valencia, CA). We genotyped all individuals at
7 nuclear, bi-parentally inherited microsatellite loci
(BM1225, BM4107, BM4208, BM6506, CSN3, Bishop
et al. 1994; RT23, RT27, Wilson et al. 1997). We conducted
polymerase chain reactions (PCR) and genotyping following
protocols described in Blanchong et al. (2006,2008). We
screened PCR products using either a Li-COR Instruments
Li-COR IR2 DNA Sequencer (NENTM, Lincoln, NE) or
Hitachi Instruments FMBIOII sequencer (Hitachi Software
Engineering Co., Ltd., Yokohama, Japan). We ran
individuals of known genotype and size standards concurrently on each gel. Two experienced laboratory personnel
scored all genotypes. We genotyped 10% of all samples a
second time for all loci as a means of quality control. Scoring
errors rates were <0.5% (1 of 231 genotypes were erroneous
during error check).
For each group of deer, we tested genotype frequencies at
all loci for Hardy–Weinberg (HW) expectations using the
exact HW test of Guo and Thompson (1992) implemented
in GENEPOP version 4.0 (Rousset 2008). We conducted
tests for evidence of genotypic linkage disequilibrium (i.e.,
whether genotypes at 1 locus are independent of genotypes at
a second locus) using the log-likelihood ratio test, with
probabilities computed based on the Markov chain method
of Raymond and Rousset (1995) implemented in GENEPOP. We used a Bonferroni correction (Rice 1989) to adjust
alpha levels for multiple tests. We calculated estimates of
Locher et al.
Deer Spatial Genetic Structure
allele frequencies and expected and observed heterozygosity
(HE, HO) for each group using MICROSATELLITE
ANALYSER v 3.12 (Dieringer and Schlötterer 2003). We
calculated group estimates of allelic richness (Ar) and the
inbreeding coefficients (FIS) using FSTAT version 2.9.3
(Goudet 2001). We derived estimates of pairwise relatedness
among individuals within each group using maximum
likelihood methods (Wagner et al. 2006) implemented in
program ML-Relate (Kalinowski et al. 2006). We selected
this method because maximum likelihood methods are
typically more accurate than other methods (Milligan 2003).
Using GENECAP (Wilberg and Dreher 2004; based on the
methods of Evett and Wier 1998), we estimated a multilocus probability of identity at the level of full siblings was
0.000498 over 7 loci.
The quality of habitat within an area occupied by an
individual can contribute to decisions to disperse (either
immigration or emigration) owing to condition or phenotype-dependent factors (Clobert et al. 2009). Hence, we
quantified the relative quality of landscape features (HSI)
associated with areas occupied by each group. We used
multiple linear regression implemented in SAS (v 9.2 SAS
Institute, Inc., Cary, NC) to associate within-group
measures of genetic diversity (HE, mean relatedness and
percentage of inter-individual pairs related at half-sibling
[HS], full-sibling [FS], or parent-offspring [PO] levels) to
HSI . We also used group size (N) and size of the MCP area
(km2) encompassing each group as predictor variables to test
for their effects. We used Bayesian methods (Foll and
Gaggiotti 2006) implemented in program GESTE to
determine the influence of mean habitat quality (HSI)
within the MCP of each group to genetic differentiation
(using point-estimates of Fst from GESTE) between the
focal group and all other groups. We used a burn-in of
50,000 replicates, a thinning interval of 20, and a sample size
of 10,000. We conducted analyses over 250,000 iterations.
Analyses of Spatial Genetic Structure Among Groups
We derived estimates of SGS (Fst, Weir and Cockerham
1984) from FSTAT and used them as a measure of SGS for
all pair-wise comparisons between the 13 groups. We used
Bonferroni corrections (Rice 1989) to adjust alpha levels to
account for multiple tests.
To test the hypothesis that the degree of SGS among
white-tailed deer in southern Michigan was a function of
geographic proximity (isolation-by-distance; Wright 1943,
Rousset 1997), we performed a Mantel test of Euclidean
distance between group centroids on Fst/(1 Fst) in
PASSaGE version 2 (Rosenberg and Anderson 2011).
We assessed statistical significance based on 1000
permutations.
Bayesian clustering methods can have difficulty identifying
contiguous genetic clusters of individuals in species and
landscape contexts characterized by isolation-by-distance
(Waples and Gaggiotti 2006, Guillot et al. 2009). However,
to further examine any strong effects of putative barriers
(e.g., major highways) that may have overwhelmed an
underlying signal of isolation-by-distance, we employed
183
Figure 1. Estimates of habitat suitability (light shade ¼ low; dark shade ¼ high), boundaries of defined deer groups (inset), and collection locations (points) of
individuals genotyped in southern Michigan, USA 1998–2000. Membership to 1 of 2 distinct genetic clusters is shown by pink circles or yellow squares. The
high-traffic highway refers to the location of Interstate-75 (bold red line).
Bayesian clustering methods that incorporated information
on individual harvest location to estimate cluster membership (e.g., Corander and Marttinen 2006, Guillot et al.
2009). We performed analyses using BAPS version 4.0
(Corander et al. 2004), implemented using the spatial genetic
mixture analysis to assign individuals to genetic clusters. We
completed 10 independent runs of K ¼ 1–13, and compared
results for consistency.
We tested for spatial genetic autocorrelation using all
individuals in GenAlEx 6.5 (Peakall and Smouse 2012) to
better understand the distances below which dispersal may be
184
restricted, and individuals are more genetically related than
expected by chance. We tested for significance using 9999
random permutations of the data, and 95% confidence
intervals; estimates of r were determined by 9999 bootstraps.
We used 10 distance classes at 20-km intervals.
Quantifying Landscape Effects on Measures of Spatial
Genetic Structure
To quantify the influence of landscape features on deer SGS,
we used cost-distance analysis implemented in ArcMap and
the multivariate optimization approach of Pérez-Espona
The Journal of Wildlife Management
79(2)
et al. (2008) to calculate the least cumulative cost of moving
through landscapes between the geometric centroids of deer
groups. We used several landscape features to calculate the
cost distance to assess their influence on deer population
structure: major interstate highways (highways classified as
high, medium, or low based on traffic volume; see above),
rivers (Michigan Center for Geographic Information 2002),
and habitat suitability (HSI). We classified habitat suitability
into low (0–35.9), intermediate (36–57.9), and high(58–100)
by optimizing the goodness of variance fit, and separated
classes using Jenk’s optimization in ArcMap, which
minimized the sum of the variance within each suitability
category (Jenks 1967). We classified suitability into the 3
categories to evaluate the specific contribution of different
measures of suitability to SGS. Because linear features
can become discontinuous when converted to 30-m raster
cells in the resistance surface, we buffered rivers, streams,
and roads by 50 m prior to converting the linear features to
30-m 30-m raster format.
We calculated least cost-distance matrices for each
landscape feature by developing several cost surfaces and
assigning each grid cell a cost ¼ 1, except for those containing
the landscape feature of interest (Pérez-Espona et al. 2008).
Considerable uncertainty exists associated with weights
assigned to landscape variables (Spear et al. 2010, Manel and
Holderegger 2013). To address this uncertainty in a
systematic and quantitative manner, for the landscape
features of interest, we assigned a range of 8 cost values
(0.0001, 0.001, 0.01, 0.1, 10, 100, 1,000, 10,000). The range
in cost values were as described in other ungulate studies
(Pérez-Espona et al. 2008). Values increased incrementally
by a power of 10 to ensure we evaluated an adequate range of
values to determine whether the features were acting as
barriers (cost > 1) or facilitating (cost < 1) gene flow. We
used Mantel tests and permutation testing (n ¼ 1,000
permutations) in Program R v 3.2.1 to determine which
cost value best quantified the effect of each landscape feature
on deer SGS, and to quantify the probability the observed
results would occur by chance. Additionally, we constructed
inter-group matrices of Euclidean distance and genetic
distance and used Mantel tests to quantify this relationship.
We determined which cost value for each landscape feature
maximized the r2 value and thus explained more of the
variability (relative to Euclidean distance alone) in
the relationship between genetic differentiation and landscape features (Pérez-Espona et al. 2008).
We then created resistance surfaces for all possible
combinations of landscape features using the cost values
for each feature that maximized r2. We ran 10 Mantel and
partial Mantel tests (i.e., partialling out Euclidean distance)
to compare which surface best explained SGS. Because of the
problem of non-independence of data points, Mantel tests
are one of the most appropriate analyses for distance-based
data (Legendre and Fortin 2010).
Because the Mantel and partial Mantel tests tend to have
high type 1 error rates (Balkenhol et al. 2009) and tend not to
correctly estimate r, we also tested the relationship between
SGS and all possible combinations (n ¼ 31) of landscape
Locher et al.
Deer Spatial Genetic Structure
features (with the optimal cost weighting) using Akaike’s
Information Criterion (AIC; Burnham and Anderson 2002)
for multiple regression models. Using multiple methods
reduced the chances of leading to erroneous methoddependent conclusions (Balkenhol et al. 2009). We
calculated AIC, relative AIC (DAIC), model likelihoods,
and Akaike weights for each model. We then used the
Akaike weights to determine parameter estimates and
unconditional standard errors (Burnham and Anderson
2002) and the best-cost model for explaining the relationship
between genetic differentiation and landscape features.
Because we compared each group with all other groups,
spatial autocorrelation in measures of inter-group differentiation may exist and affect evaluations of model fit via AIC
(Goldberg and Waits 2010). Thus, we used multiple tests
(i.e., Mantel, partial Mantel, AIC) used for analyses to
combat shortcomings of any single method.
RESULTS
Spatial Heterogeneity in White-Tailed Deer Habitat
Suitability
The quality of deer habitat as reflected in measures of habitat
suitability (HSI) varied widely at multiple spatial scales
across the study area (Fig. 1) and within areas occupied by
each group of deer (Table 1). The central portion of the study
area, generally extending from southwest to northeast, was
predominantly agricultural lands. Mean HSI estimates for
white-tailed deer groups collected from these areas (e.g.,
groups 2, 6, 12, and 13) were considerably lower than for
groups collected in peripheral areas (Table 1). Large
contiguous areas of crops were interspersed with comparatively smaller patches composed of small woodlots and
woody vegetation associated with wind-rows and riparian
corridors as well as highways. Areas to the northwest, south,
and east of this central region were composed of higher
proportions of woody vegetation and were of comparatively
higher habitat quality. The entire study area was bisected by
several major highways and rivers (Fig. 1).
Measures of Genetic Diversity Within Groups
Genotypic frequencies at all loci within each white-tailed
deer group conformed to HW expectations, and we did not
detect any evidence of genotypic disequilibrium. Mean
estimates of group genetic diversity varied among groups
(allelic richness, Ar: 6.58–9.46, observed heterozygosity HO:
0.761–0.887, expected heterozygosity HE: 0.797–0.845,
Wright’s inbreeding coefficient Fis: 0.089 to 0.103;
Table 1). We did not observe latitudinal or longitudinal
trends in measures of diversity among groups (P > 0.05).
Mean habitat quality (HSI) within group sampling areas did
not predict group measures of genetic diversity (HE, mean
relatedness); however, we found a significant positive
relationship between mean group relatedness and group
sample size (r2 ¼ 0.67, P < 0.01). Collectively, measures of
within group genetic diversity were associated with the
sample size and/or area but not the quality of habitat within
the area encompassing groups.
185
Table 1. Summary measures of genetic diversity including estimates of inter-individual relatedness within 13 groups of white-tailed deer sampled in
southern Michigan, 1998–2000, minimum convex polygon estimates of areas (km2) encompassing all individuals within a group, and mean habitat suitability
indices (HSI) in areas encompassing each group. We measured HSI on a scale of 0–100; 100 represents optimal conditions.
Measures of genetic diversitya
Group
k
A
HO
HE
Fis
Mean relatedness
Group area
Mean HSI/group area
8.86
10.14
8.14
8.71
10.71
8.43
8.86
10.43
10.43
9.57
11.14
10.71
9.57
9.085
7.072
9.463
6.765
7.648
7.916
8.402
7.554
6.577
8.754
8.611
7.669
8.277
0.887
0.830
0.810
0.787
0.817
0.837
0.809
0.818
0.761
0.864
0.830
0.844
0.807
0.817
0.844
0.831
0.849
0.817
0.844
0.813
0.815
0.797
0.838
0.845
0.833
0.815
0.089
0.017
0.026
0.103
0.001
0.009
0.005
0.003
0.045
0.021
0.018
0.026
0.023
0.0456
0.0556
0.0341
0.0376
0.0525
0.0384
0.0369
0.0504
0.0524
0.0470
0.0526
0.0535
0.0305
663.02
1,277.16
259.66
390.61
978.13
541.23
659.72
702.65
808.06
342.59
1,129.76
672.77
1,116.53
48.30
34.06
46.19
49.75
45.13
29.28
46.45
59.59
49.83
46.34
53.34
35.97
37.25
Sample size
1
2
3
4
5
6
7
8
9
10
11
12
13
19
30
13
17
38
14
16
26
38
19
42
31
19
a
k, mean number of alleles; A, allelic richness; HO, observed heterozygosity; HE, expected heterozygosity; Fis, Wright’s inbreeding coefficient.
Statistical significance of Fis (P < 0.05).
Analyses of Spatial Genetic Structure
Genetic differentiation over the entire study was significant
(mean inter-group Fst ¼ 0.019, P < 0.01, range 0.000–0.053;
Table 2). Deer sampled in the eastern portion of the study
area (groups 9 through 11; Fig. 1) were particularly divergent
from deer across the central, northern, western, and southern
portion of the study area (Table 2). Inter-group genetic
differentiation increased when inter-group geographical
distance increased (r2 ¼ 0.186, P < 0.05; Fig. 2).
At the level of individuals, Bayesian clustering revealed
evidence for 2 genetic clusters. One cluster consisted of
individuals primarily centered in the eastern region of the
study area predominantly associated with groups 9 through
11 (Fig. 1). The other genetic cluster consisted of
individuals that were more widely distributed throughout
the south, east, and northern regions of the study area.
Individual relatedness among individuals was significantly spatially autocorrelated over distances within the
boundaries of the sampling groups (Fig. 3). However,
evidence for gene flow over geographically more expansive
areas (Fig. 2) suggests that multi-generational gene flow or
long-distance gene flow is also important to SGS at larger
spatial extents.
Effects of Within Patch Habitat Quality on SGS
Deer collected from groups inhabiting areas of lowest and
highest habitat quality as reflected by mean group HSI were
less genetically differentiated from other groups than
groups of deer inhabiting areas of intermediate quality
(Fig. 4). However, estimated group HSI was not a
significant predictor of SGS (P > 0.05), likely owing to
the non-linear nature of the relationship between HSI and
inter-group Fst values. The addition of a quadratic term in
the regression of inter-population genetic variation and
mean population HSI (r2 ¼ 0.436; P < 0.05) had a better fit
than the linear model (r2 ¼ 0.013; P > 0.05) suggesting that
the relationship between patch habitat quality and
propensity for gene flow were neither linear nor positive
across the range of the mean HSI values of the 13 sampled
groups. This result was consistent with analyses of high,
intermediate, and low HSI and SGS (see below). Mean
inter-group Fst differed among groups (Table 2). For
Table 2. Pair-wise genetic differentiation (Fst; above diagonal) and Euclidean distance (km; below diagonal) between white-tailed deer sampled from 13
groups in southern Michigan, 1998–2000. Values with an asterisk (*) indicate pairwise Fst values that are significantly different from (P < 0.05).
White-tailed deer groups
1
1
2
3
4
5
6
7
8
9
10
11
12
13
186
2
0.0115
70.53
92.44
115.53
128.7
102.73
151.82
134.57
116.78
89.91
65.14
49.05
95.04
35.42
45.3
76.43
73.39
117.49
126.62
151.55
137.48
122.96
64.33
102.69
3
0.0204
0.013
37.68
99.76
105.84
145.58
160.64
186.07
169.83
152.31
98.35
138.11
4
0.0127
0.0185
0.0152
72.13
91.03
121.02
146.9
188.67
179
167.04
105
133.92
5
6
*
0.0165
0.0103
0.0291
0.0291*
39.15
49.11
84.74
148.8
152.31
152.82
89.95
89.67
7
*
0.0224
0.0071
0.0117
0.0192*
0.0099
49.47
56.03
110.01
113.49
116.13
64.33
50.82
0.0204
0.0151
0.0325
0.0368*
0
0.0164
52.09
131.68
146.15
156.17
104.8
79.28
8
0.0105
0.0121*
0.0309*
0.0269*
0.0122
0.0142
0.013
81.78
101.92
118.56
85.96
41.76
9
*
0.0286
0.0493*
0.042
0.009
0.0444*
0.0469*
0.0528*
0.0436*
34.47
63.66
87.77
59.19
10
11
12
13
0.0299
0.0347*
0.0279*
0.0056
0.0435*
0.0362*
0.0498*
0.041*
0.006
0.0093
0.0134
0.0092
0.0074*
0.0304*
0.0248*
0.0265*
0.0153*
0.0213*
0.0100*
0.0014
0.0033
0.0099
0.0131*
0.0073
0.0038
0.0125
0.0132*
0.0350*
0.0296*
0.0137*
0.0039
0.0086
0.0181
0.016
0.0082
0.0105
0.0012
0
0.0322*
0.0231
0.0043
0.0031
29.89
74.01
67.03
65.16
78.49
48.52
The Journal of Wildlife Management
79(2)
Figure 2. Analysis of isolation by distance showing the relationship between inter-group genetic differentiation as a function of distance (R2 ¼ 0.186; P < 0.05)
for white-tailed deer in southern Michigan (USA) harvested between 1998–2000.
example, groups located in the center of the study area
inhabiting areas of generally poorer habitat (groups 12 and
13) and peripheral groups with comparatively higher HSI
(e.g., groups 1 and 11) were characterized by lower and less
variable inter-group Fst values (Fig. 4).
Effects of the Landscape Matrix on Spatial Genetic
Structure
Four landscape features explained more of the variability
in genetic differentiation among groups, as indicated by a
higher r2 value than Euclidean distance alone. The hightraffic highway (i.e., Interstate highway 75), rivers, and
intermediate habitat suitability appeared to act as barriers
to gene flow, and the cost values maximizing r2 for these
features were 10,000, 1,000, and 100, respectively
(Table 3). Low habitat suitability facilitated gene flow;
the cost value that maximized r2 was less than 1
(estimated resistance weight 0.1). Medium- and
low-traffic highways and areas classified as high habitat
suitability did not explain more of the variation in genetic
differentiation than Euclidean distance alone, and thus
we did not include them in final resistance surfaces to
explain SGS.
The Mantel and partial Mantel tests indicated that the
best models for explaining genetic differentiation included
rivers, intermediate habitat suitability, and low habitat
suitability (Table 4). However, the confidence intervals for
0.050
0.040
0.030
r
0.020
0.010
0.000
-0.010
-0.020
-0.030
0
20
40
60
80
100
120
140
160
180
Distance class (start point)
Figure 3. Correlograms of correlation coefficients (r) of geographic and genetic distance at variable distance classes (km) for individual white-tailed deer
harvested in southern Michigan, USA (1998–2000). Upper and lower error bars are bound by 95% confidence intervals around each r, and dashed lines indicated
95% confidence limits (upper [U] and lower [L]) around the null hypothesis of a random spatial distribution of genotypes.
Locher et al.
Deer Spatial Genetic Structure
187
Figure 4. Graphical relationship between mean genetic differentiation (natural log of point estimates over all pairwise comparisons) between groups of whitetailed deer harvested in southern Michigan, USA (1998–2000) and average habitat suitability index (HSI) estimates for the sampling area encompassing each
white-tailed deer group.
the cost value (cost ¼ 100) maximizing r2 (r2 ¼ 0.555,
lower ¼ 0.33, upper ¼ 0.71) for intermediate habitat suitability overlapped the confidence limits for the r2 value at a
cost of 0.1, thus suggesting that the inhibiting relationship
to genetic differentiation may be weak (Table 3). Likewise,
the confidence limits for the cost value (cost ¼ 0.1)
maximizing r2 (r2 ¼ 0.430, lower ¼ 0.31, upper ¼ 0.57) for
low habitat suitability overlapped that of the r2 value for a
cost of 10, thus suggesting that the facilitating relationship
to genetic differentiation may be weak (Table 4). The
second-best model according to the Mantel and partial
Mantel tests included the high-traffic highway and rivers.
The lower confidence limit of the maximum r2 value for the
high-traffic highway was greater than the upper limit for
Table 3. Estimates of r2 from Mantel tests among measures of inter-group genetic differentiation [Fst/(1 Fst)] between groups of white-tailed deer
harvested in southern Michigan, USA (1998–2000) and each of several landscape features assessed at each of several resistance values. Values with a dagger
(†) indicate features and costs that improved model predictability over Euclidean distance alone (r2 ¼ 0.381). Upper and lower 95% confidence limits are
given in parentheses below significant cost values (P < 0.05) with higher r2 values than Euclidean distance. Hwy, highway; HSI, habitat suitability index.
Cost values
Facilitate gene flow
All hwy (same weights)
High traffic hwy
(I-75)†
Medium traffic
hwy
Low traffic
hwy
Rivers†
High HSI
Intermediate
HSI†
Low HSI†
Euclidean distance
*
Inhibit gene flow
0.0001
0.001
0.01
0.1
10
100
1,000
10,000
0.375
0.213
0.243
0.216* (0.13 – 0.34)
0.175
0.375
0.213*
(0.12 – 0.33)
0.176
0.182
0.241
0.243*
(0.15 – 0.37)
0.239
0.375
0.382*
(0.26 – 0.52)
0.381
0.369
0.424*
(0.30 – 0.57)
0.341
0.226
0.593*
(0.46 – 0.74)
0.131
0.151
0.631†*
(0.47 – 0.74)
0.138
0.231
0.254
0.297
0.318
0.381
0.375
0.319
0.237
–0.133
–0.133
–0.125
0.000
0.02
–0.101
0.031
0.043
0.307
0.294
0.32
0.111
0.326*
(0.22 – 0.43)
0.430†*
(0.31 – 0.57)
0.372*
(0.27 – 0.52)
0.326
0.555†*
(0.33 – 0.71)
–0.057
0.396†*
(0.25 – 0.49)
0.029
0.547*
(0.08 – 0.84)
–0.168
0.382
–0.013
–0.147
0.378*
(0.26 – 0.51)
0.361
0.399*
(0.28 – 0.54)
0.352*
(0.23 – 0.49)
–0.125
0.506*
(0.50 – 0.78)
–0.143
0.381
(0.27 – 0.51)
Indicates significant cost values (P < 0.05).
188
The Journal of Wildlife Management
79(2)
Table 4. Correlation coefficients, lower and upper confidence limits from Mantel and partial Mantel tests on the relationship between resistance surfaces
containing various landscape features and spatial genetic structure [Fst/(1 Fst)] among white-tailed deer groups in southern Michigan, 1998–2000. Hwy,
high traffic highway (Interstate highway I-75); HSI, habitat suitability index; EucDist, Euclidean distance.
Mantel
Features
River, medium_HSI, low_HSI
Hwy, river
Medium_HSI, low_HSI
River, medium_HSI
Hwy, river, low_HSI
Hwy, river, medium_HSI
Hwy, river, medium_HSI, low_HSI
Hwy, low_HSI
Hwy, medium_HSI
EucDist
2
r
Lower
0.712
0.673
0.652
0.639
0.632
0.629
0.626
0.624
0.615
0.381
0.629
0.521
0.509
0.541
0.477
0.495
0.485
0.464
0.469
0.266
cost values <1. For rivers, cost values <1 were not
significant. Thus, the presence or absence of rivers in the
landscape matrix between populations was predictive of
inter-group Fst (r2 ¼ 0.396) and significantly exceeded
relationships based on Euclidean distance alone (r2 ¼ 0.381;
Table 3). The strongest inhibitor of gene flow (i.e.,
significant improvement over the model with Euclidean
distance alone) was when we conducted the cost analysis
using the high-traffic highway (i.e., Interstate highway 75;
Fig. 1) that separates groups 9 through 11 from the rest of
the study area. This highway also separated the majority of
members of the 2 inferred genetic clusters (see above;
Fig. 1).
According to AIC, the final model of best fit included the
high-traffic highway, rivers, intermediate habitat suitability,
and Euclidean Distance (AIC ¼ 704.7, r2 ¼ 0.426; Table 5), and also explained more of the variation in genetic
differentiation than Euclidean distance alone (r2 ¼ 0.186).
The second best model contained the high-traffic highway
Partial Mantel
Upper
2
r
Lower
Upper
0.793
0.833
0.774
0.763
0.805
0.841
0.831
0.813
0.829
0.513
0.652
0.602
0.584
0.586
0.554
0.544
0.546
0.544
0.528
0.563
0.418
0.41
0.466
0.383
0.399
0.395
0.36
0.352
0.731
0.788
0.722
0.719
0.76
0.783
0.777
0.757
0.77
and rivers (r2 ¼ 0.384) and also explained more of the genetic
variation than Euclidean distance alone. Permutation testing
revealed that these relationships did not occur by random
chance (P < 0.001). Final model averaged parameter
estimates were b^0 ¼ 0:004(SE ¼ 0.003), b^1 ¼ 2:65E 9
(SE ¼ 7.01E10),
b^2 ¼ 1:60E 7
(SE ¼ 5.13E8),
b^3 ¼ 7:39E 8(SE ¼ 4.12E8), b^4 ¼ 2:29E 7(SE ¼ 1.01
E7) for high traffic highways, rivers, medium HSI, and
Euclidean distance, respectively (Table 5).
DISCUSSION
Evaluation of landscape features associated with connectivity
between patches has wide applications in wildlife conservation planning at multiple spatial scales (Calabrese and Fagan
2004, Urban et al. 2009). We developed a connectivity model
for white-tailed deer and evaluated multiple least cost models
to measure associations between landscape features and
measures of genetic diversity within and among deer groups
Table 5. Akaike’s Information Criterion (AIC), relative AIC (DAIC) values, model likelihood, and Akaike weights (wt) for various models explaining
spatial genetic structure [Fst/(1 Fst)] among white-tailed deer groups in southern Michigan, 1998–2000. Hwy, high traffic highway (Interstate highway
I-75); Low_HSI, low habitat suitability; Med_HSI, medium habitat suitability; Hi_HSI, high habitat suitability; EucDist, Euclidean distance.
Parameter
Hwy, river, med_HSI, EucDist
Hwy, river
Hwy, river, low_HSI
Hwy, river, med_HSI
Hwy, river, low_HSI, med_HSI
Hwy, med_HSI
Hwy, low_HSI
Hwy, low_HSI, med_HSI
Hwy
River, med_HSI
Med_HSI
Low_HSI, med_HSI
River, low_HSI, med_HSI
Low_HSI
River, low_HSI
River
EucDist
Locher et al.
Deer Spatial Genetic Structure
R2
AIC
DAIC
Likelihood
Akaike wt
0.426
0.384
0.387
0.385
0.393
0.347
0.342
0.351
0.297
0.270
0.244
0.270
0.253
0.239
0.222
0.220
0.186
704.7
703.2
701.6
701.4
700.4
698.7
698.1
697.1
694.9
690.0
689.2
688.2
688.0
687.0
686.8
686.8
683.6
0
1.5
3.1
3.3
4.3
6.0
6.6
7.6
9.8
14.7
15.5
16.5
16.7
17.7
17.9
17.9
21.1
1.00
0.47
0.21
0.19
0.12
0.05
0.04
0.02
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.47
0.22
0.10
0.09
0.06
0.02
0.02
0.01
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
189
in southern Michigan. Our global analysis using multiple
methods (Mantel tests and AIC model evaluation) to
quantify effects of factors influencing SGS provided evidence
that landscape features between groups, such as rivers and
high-traffic highways, were barriers to movement and were
significant predictors of SGS (Tables 3 and 5). Natural
landscape features such as rivers have previously been
documented as imposing a barrier to white-tailed deer gene
flow (Blanchong et al. 2008, Robinson et al. 2012). Manmade features of more recent (<65 years) origin such as
highways were also found to inhibit gene flow (Long et al.
2010) although white-tailed deer adapt well to urbanization
and fragmentation (Ditchkoff et al. 2006). The largest
(highest traffic volume) and oldest highway (Interstate
highway 75 in the eastern portion of the study area
constructed in the mid 1950–1960s) represented a significant
impediment to gene flow (Table 3), and coincided with the
boundaries of the 2 genetic clusters (Fig. 1). Roads
have frequently been shown to have negative effects on
wildlife movement (Foreman and Alexander 1998,
Trombulak and Frissell 2000, Underhill and Angold
2000, Balkenhol and Waits 2009). With increasing human
populations, transportation networks and highway traffic
volume are likely to increase with expansion of urban and
suburban areas (Borda-de-Aqua et al. 2011). Given the
relationships between SGS and presence of roads for highly
vagile white-tailed deer, spatial genetic structure of species
with lower vagility and lower variance effective population
size would be expected to be more dramatically affected by
anthropogenic disturbance.
Interpreting results from statistical tests of spatially
distributed phenomena is challenging because of issues
with spatial autocorrelation and strengths and weaknesses of
each test. For instance, partial Mantel tests and their
associated analyses (e.g., Multiple Regression on Distance
Matrices, MRM) have high type 1 error rates (Balkenhol
et al. 2009, Legendre and Fortin 2010), and information
theoretic approaches assume data independence (Burnham
and Anderson 2002). However, landscape genetic studies
commonly employ a number of analytical methods, and the
challenge is then to interpret several different model
results while taking into account the limitations of the
approaches. Both of our modeling approaches converge on
the results that rivers and high-traffic highways were
significant predictors of SGS. The relationship between
genetic diversity and landscape-level habitat suitability
metrics was less obvious. However, measures of habitat
quality quantified within and between areas occupied by deer
groups were also associated with SGS, as exemplified by
regression analysis of HSI for sampling areas encompassing
white-tailed deer groups and Fst (Fig. 4) and the results of
Mantel tests (Table 3), but to a lesser degree than rivers and
the high-traffic highway. For example, groups collected from
areas where habitat was of intermediate quality were more
highly differentiated from deer in other groups relative to
deer collected from areas of high and low habitat quality
(Fig. 4). Thus, the permeability of intermediate quality
habitat may be lower, meaning these areas inhibit gene flow,
190
whereas areas of low habitat suitability may facilitate gene
flow (Tables 3 and 5). In addition to the comparison of mean
genetic differentiation and HSI, model selection based on
AIC suggested that intermediate habitat suitability (inhibitor) and low habitat suitability (facilitator) were significantly
better predictors of genetic differentiation than Euclidean
distance alone. Although the Mantel tests suggested these
habitat relationships may not be significant (Table 3), the
partial Mantel test and model selection based on AIC alluded
to their significance (Tables 4 and 5). Alternatively, the
habitat classes may not correctly reflect permeability or gene
flow, or may not be sufficient to detect patterns at the scale of
analysis. For instance, the variability in HSI is diluted as scale
becomes coarser. The variation in significance of habitat
suitability with different analysis methods warrants further
investigation of the effects of habitat on spatial genetic
structure at large landscape scales.
One important assumption underlying the approaches used
in this study, and those commonly used in landscape genetic
analyses (Hall and Beissinger 2014), is that the relationships
of interest are stationary or homogeneous spatially. The nonrandom distributions of habitats of comparatively higher and
lower quality across the study area (Fig. 1) and the prominent
effects of interstate highway 75 on SGS (i.e., based on the
abrupt discontinuity in genetic cluster; Fig. 1) suggests nonstationarity of effects of landscape features and barriers
affecting SGS. Alternative analytical methods such as
geographically weighted regression (GWR; Fotheringham
et al. 2002) can account for spatial components of the data
including local regression. However, GWR analyses requires
large datasets (ideally at least 40 measurements in weighted
regressions for each set of points (distance intervals;
Fotheringham et al. 2012), and for this reason, GWR has
not often been applied in landscape genetic studies. Across
our study area, the number of groups and inter-group
comparisons within and between patches were not sufficient
to use this approach. Other approaches such as multiple
regression on distance matrices (MRM; Lichstein 2007)
would allow the examination of how landscape features affect
patterns of genetic variance at varying distances; however, the
restricted number of groups precluded use of this or other
similar approaches.
The low inter-group genetic differentiation in groups of
deer inhabiting lower quality habitats may be due to lower
deer density in these areas, especially if the low suitability
areas were isolated or surrounded by intermediate habitat
quality, which may inhibit gene flow (Table 3). For example,
Roseberry and Woolf (1998) reported a positive curvilinear
relationship between white-tailed deer densities and habitat
suitability in an agricultural region in the mid-western
United States similar to our study area. Genetic differentiation may be low in areas of higher quality habitat if deer were
able to meet life requisites and therefore, were less likely to
disperse (King 1938, Clobert et al. 2004), particularly if
immigration into these areas was also concurrently low. For
instance, Felix et al. (2007) found that non-migratory deer in
Michigan occupied areas of comparatively higher quality
winter and summer habitat than migratory deer that
The Journal of Wildlife Management
79(2)
occupied areas where at least 1 critical habitat component
was missing from their seasonal home range. Deer in
Pennsylvania dispersed greater distances in areas with less
forest cover, which was related to lower habitat quality (Long
et al. 2005). Social organization of white-tailed deer (i.e.,
differences in degree of male and female philopatry) may also
explain SGS (Chesser 1991, Scribner et al. 2001). To
substantiate theses hypotheses, future research should
investigate the interrelationships among habitat suitability,
deer densities, movement patterns including dispersal and
migration, social organization, and SGS.
Evidence for multi-generational or long-distance dispersal
in our study is based on the dispersion of individuals with
high posterior probabilities of affiliation with the genetic
cluster whose members are predominately found in the
eastern portion of the study area (Fig. 1) and the significance
associated between inter-group genetic differentiation and
geographic distance across the entire study area (Fig. 2).
Although the effects of habitat suitability on long-distance
movements are unknown, long-distance movement of whitetailed deer does occur. For instance, in our study area,
distances up to 60 km approximates the extent of detectable
positive genetic structure. In upper Michigan, seasonal
movements of greater than 50 km have been documented
(Verme 1973). Although the presence of the highway likely
impedes movement, underpasses and natural drainages or
waterways may facilitate deer passage in some instances
(Donaldson 2007, Corlatti et al. 2009, Ford et al. 2009).
Although quantitative data pertaining to population
density are not available, density certainly varies across the
large southern Michigan study area and throughout the
course of a year and could affect interpretations of results.
Spatial and temporal variability in deer density is likely due in
part to habitat quality and thus we expected collinearity
between site- (or group-) specific density and habitat quality.
For example, Roseberry and Woolf (1998) found that 81% of
the variation in local deer density was attributed to habitat
quality. Density likely also varies spatially and temporally in a
sex-specific manner (Bowyer 2004). Dispersal, migration,
and human activities such as baiting during hunting season
would also affect densities in localized areas as well as SGS
(Blanchong et al. 2006). Samples were collected during the
annual firearms harvest period but no data are available on
hunter effort or method of harvest. Movements of deer and
changes in local densities throughout the course of the year
may be a source of variability in our data. However, the fact
that the results showed relationships between measures of
SGS, barriers, and HSI in the absence of information on
variation in density suggests the results are likely robust.
Our analyses relied on measures of potential resistance of
landscape features to deer gene flow. Development of
predictive measures of resistance is challenging. Dispersal is
frequently conditional upon resource availability (e.g., food,
access to mates), and population density (Ronce et al. 2001,
Clobert et al. 2009), which vary spatially and temporally.
Resource availability can be estimated with measures of
habitat suitability, which are generalized estimates not
reflecting inherent variation at the microhabitat scale or may
Locher et al.
Deer Spatial Genetic Structure
vary because of factors such as drought, flooding, crop
rotations, or other events. Spatial and temporal scales over
which samples are collected (Anderson et al. 2010) and cost
surfaces (Spear et al. 2010) associated with landscape features
should be considered in experimental designs. Our estimates
of habitat suitability were based on data independent of
this study using telemetry data on deer in several portions of
our study area in southern Michigan (Pusateri 2003, Hiller
2007). Generalizations of the relative effects of habitat
features on SGS documented in Michigan to other areas of
the species range will require additional study.
Weighting of landscape features for the development of the
least cost distances is based on perceived measures of
permeability. Relative cost values assigned to different
landscape features are usually arbitrary because the true cost
to wildlife is unknown (Pérez-Espona et al. 2008). The cost
value assigned to each landscape feature will also affect the
modeled relationship between genetic structure and landscape features (Spear et al. 2010, Manel and
Holderegger 2013). Our approach of evaluating model fit
using a range of cost values for each feature and identifying
the weighting value that improved model fit associated with
variation in the relationship between genetic structure and
landscape features explicitly addresses the question of
uncertainty in a systematic manner, and allowed us to
identify relative influences of landscape features on genetic
structure of white-tailed deer in Michigan.
MANAGEMENT IMPLICATIONS
Evaluating impacts of anthropogenic activities on natural
systems is becoming increasingly critical as human populations are expanding and affecting animal movements and
habitat occupancy. Studies documenting the effects of
landscape features on gene flow in highly mobile species
of conservation and management interest are limited,
particularly in human-altered landscapes. The relative
impacts of natural and man-made landscape features on
gene flow in white-tailed deer speak to the importance of
joint ecological and genetic analyses in present and future
conservation efforts of landscapes and wildlife species using
them. We quantified the importance of roads and rivers on
gene flow of white-tailed deer. Based on our data, these
predominant features of human-modified landscapes are
barriers to gene flow and can be used in delineating
management units and to direct efforts to control immigration into and emigration from areas bounded by these major
barriers. Future studies quantifying the influence of
landscape features on wildlife, both within and between
locales will help managers and planners design landscapes to
meet human needs while understanding potential implications of anthropogenic activities on wildlife.
ACKNOWLEDGMENTS
Funding for this project was provided by the Wildlife
Division of the Michigan Department of Natural Resources
and through the Partnership for Ecosystem Research and
Management (PERM) cooperative program between the
Michigan Department of Natural Resources andthe Depart191
ment of Fisheries and Wildlife at Michigan State University.
Support was also provided through the School of Forest
Resources at the University of Arkansas at Monticello and the
Statistical Consulting Center at Grand Valley State
University. B. Smyth and K. Filcek provided valuable assistance
in the lab and with data analysis. We thank R. Holderegger, J.
Pusateri, B. Rudolf, and members of the Scribner lab for
comments that greatly improved the manuscript.
LITERATURE CITED
Aebischer, N. J., P. A. Robertson, and R. E. Kenward. 1993. Compositional
analysis of habitat use from animal radiotracking data. Ecology 74:
1313–1325.
Anderson, C., B. K. Epperson, M.-J. Fortin, R. Holderegger, P. James, M.
Rosenberg, K. T. Scribner, and S. Spear. 2010. Considering spatial and
temporal scale in landscape-genetic studies of gene flow. Molecular
Ecology 19:3565–3575.
Andreasen, J. K., P. V. O’Neill, and R. Noss. 2001. Considerations for the
development of a terrestrial index of ecological integrity. Ecological
Indicators 1:21–35.
Balkenhol, N., and L. P. Waits. 2009. Molecular road ecology: exploring the
potential of genetics for investigating transportation impacts on wildlife.
Molecular Ecology 18:4151–4164.
Beier, P., D. Majka, and W. D. Spencer. 2008. Forks in the road: choices in
procedures for designing wildland linkages. Conservation Biology 22:836–851.
Beyer, H. L. 2004. Hawth’s Analysis Tools for ArcGIS. http://www.
spatialecology.com/htools. Accessed 12 Aug 2008.
Bishop, M. D., S. M. Kappes, J. W. Keele, R. T. Stone, S. L. F. Sunden,
G. A. Hawkins, S. S. Toldo, R. Fries, M. D. Grosz, and J. Yoo. 1994. A
genetic linkage map for cattle. Genetics 136:619–639.
Blanchong, J. A., K. T. Scribner, A. N. Kravchenko, and S. R. Winterstein.
2007. A genealogical basis for disease infection risk in free-ranging
wildlife. Biological Letters 2007:103–107.
Blanchong, J. A., K. T. Scribner, B. K. Epperson, and S. R. Winterstein.
2006. Changes in artificial feeding regulations impact white-tailed deer
microgeographic genetic structure. Journal of Wildlife Management
70:1037–1043.
Blanchong, J. A., M. D. Samuel, K. T. Scribner, B. Weckworth, J.
Langenberg, and K. Filcek. 2008. Landscape genetics and the spatial
distribution of chronic wasting disease. Biological Letters 4:130–133.
Blanchong, J. A., A. B. Sorin, and K. T. Scribner. 2013. Genetic diversity
and population structure in suburban white-tailed deer. Journal of Wildlife
Management 77:855–862.
Boulet, M., S. Couturier, and S. Côté. 2007. Integrative use of spatial,
genetic, and demographic analyses for investigating genetic connectivity
between migratory, montane, and sedentary caribou herds. Molecular
Ecology 16:4223–4240.
Borda-de-Aqua, L., L. Navarro, C. Gavinhos, and H. M. Pereira. 2011.
Spatio-temporal impacts of roads on the persistence of populations:
analytic and numerical approaches. European Journal of Wildlife
Resources 55:33–43.
Bowyer, R. T. 2004. Sexual segregation in ruminants: definitions,
hypotheses, and implications for conservation and management. Journal
of Mammalogy 85:1039–1052.
Bowler, D. E., and T. G. Benton. 2005. Causes and consequences of animal
dispersal strategies: relating individual behaviour to spatial dynamics.
Biological Review 80:205–225.
Burnham, K. P., and D. R. Anderson. 2002. Model selection and
multimodel inference: a practical information-theoretic approach. Second
edition. Springer, New York, New York, USA.
Calabrese, J. M., and W. F. Fagan. 2004. A comparison-shopper’s guide to
connectivity matrices. Frontiers in Ecology and the Environment 2:529–536.
Chesser, R. K. 1991. Influence of gene flow and breeding structure on gene
diversity within populations. Genetics 129:573–583.
Chietkiewicz, C. B., and M. S. Boyce. 2009. Use of resource selection
functions to identify conservation corridors. Journal of Applied Ecology
18:561–573.
Clobert, J., R. S. Ims, and F. Rousset. 2004. Causes, mechanisms and
consequences of dispersal. Pages 307–335 in I. Hanski and O. E. Gagiotti,
192
editors. Ecology, genetics and evolution of metapopulations. London,
England, Elsevier, Academic Press.
Clobert, J., J. F. Le Galliard, and J. Cote. 2009. Informed dispersal,
heterogeneity in animal dispersal syndromes and the dynamics of spatially
structured populations. Ecology Letters 12:197–209.
Corander, J., P. Waldmann, and P. Marttinen. 2004. BAPS 2: Enhanced
possibilities for the analysis of genetic population structure. Bioinformatics
20:2363–2369.
Corander, J., and P. Marttinen. 2006. Bayesian identification of admixture
events using multi-locus molecular markers. Molecular Ecology 15:
2833–2843.
Coulon, A., G. Guillot, and J.-F. Cosson. 2006. Genetic structure is
influenced by landscape features: empirical evidence from a roe deer
population. Molecular Ecology 15:1669–1679.
Corlatti, L., K. Hacklader, and F. Frey-Roos. 2009. Ability of wildlife
overpasses to provide connectivity and prevent genetic isolation.
Conservation Biology 23:548–556.
Côté, S. D., T. P. Rooney, J.-P. Tremblay, C. Dussault, and D. M. Waller.
2004. Ecological impacts of deer overabundance. Annual Reviews in
Ecology, Evolution, and Systematics 35:113–147.
Crooks K. R., and M. A. Sanjayan. 2006. Connectivity conservation:
maintaining connections for nature. Cambridge University Press, Cambridge, United Kingdom.
Cushman, S. A., and E. L. Landguth. 2010. Spurious correlations and
inference in landscape genetics. Molecular Ecology 19:3592–3602.
Cushman, S. A., K. S. McKelvey, J. Hayden, and M. K. Schwartz. 2006.
Gene flow in complex landscapes: testing multiple hypotheses with casual
modeling. American Naturalist 168:486–499.
DeMita, S., A.-C. Thuillet, L. Gay, N. Ahmadi, S. Manel, J. Ronfort, and
Y. Vigouroux. 2013. Detecting selection along environmental gradients:
analysis of eight methods and their effectivelness for outbreeding and
selfing population. Molecular Ecology 22:1383–1399.
Diefenbach, D. R., E. S. Long, C. S. Rosenberry, B. D. Wallingford, and
D. R. Smith. 2008. Modeling distribution of dispersal distances in male
white-tailed deer. Journal of Wildlife Management 72:293–299.
Dieringer, D., and C. Schlötterer. 2003. MICROSATELLITE ANALYSER (MSA): a platform independent analysis tool for large microsatellite
data sets. Molecular Ecology Notes 3:167–169.
Diniz-Filho, J. A. F., T. F. Rangel, and L. M. Bini. 2008. Model selection
and information theory in geographical ecology. Global Ecology and
Biogeography 17:479–488.
Ditchkoff, S. S., S. T. Saalfeld, and C. J. Gibson. 2006. Animal behavior in
urban ecosystems: modifications due to human-induced stress. Urban
Ecosystems 9:5–12.
Donaldson, B. 2007. The use of highway underpasses by large mammals and
other wildlife in Virginia: factors influencing their effectiveness.
Environmental Issues 2011:157–164.
Dungan, J. L., J. N. Perry, M. R. T. Dale, P. Legendre, S. Citron-Pousty,
M.-J. Fortin, A. Jakomulska, M. Miriti, and M. S. Rosenberg. 2002. A
balanced view of scale in spatial statistical analysis. Ecography 25:626–640.
Epperson, B. K. 1993. Recent advances in correlation studies of spatial
patterns of genetic-variation. Evolutionary Biology 27:95–155.
Epps, C. W., J. D. Wehausen, V. C. Bleich, S. G. Torres, and J. S.
Brashares. 2007. Optimizing dispersal and corridor models using
landscape genetics. Journal of Applied Ecology 44:714–724.
Fahrig, L. 2007. Non-optimal animal movement in human-altered
landscapes. Functional Ecology 21:1003–1015.
Felix, A. B., D. P. Walsh, B. D. Hughey, H. Campa, and S. R. Winterstein.
2007. Applying landscape-scale habitat-potential models to understand
deer spatial structure and movement patterns. Journal of Wildlife
Management 71:804–810.
Foll, M., and O. E. Gaggiotti. 2006. Identifying the environmental factors that
determine the genetic structure of populations. Genetics 174:875–891.
Ford, A. T., A. P. Clevenger, and A. Bennett. 2009. Comparison of methods
of monitoring wildlife crossing-structures on highways. Journal of
Wildlife Management 73:1213–1222.
Forman, R. T. T., and L. E. Alexander. 1998. Roads and their major
ecological effects. Annual Review Ecology Evolution and Systematics
29:207–231.
Fotheringham, A. S., C. Brunsdon, and M. E. Charlton. 2002.
Geographically weighted regression: the analysis of spatially varying
relationships. Wiley, New York, New York, USA.
The Journal of Wildlife Management
79(2)
Geffen, E., M. J. Anderson, and R. K. Wayne. 2004. Climate and habitat
bariers to dispersal in the highly mobile grey wolf. Molecular Ecology
13:2481–2490.
Gladfelter, H. L. 1984. Whitetail populations and habitats; Midwest
agricultural region. Pages 427–440 in L. K. Halls , editor. White-tailed
deer ecology and management. Stackpole Books, Harrisburg,
Pennsylvania, USA.
Goldberg, C. S., and L. P. Waits. 2010. Comparative landscape genetics of
two pond-breeding amphibian species in a highly agricultural landscape.
Molecular Ecology 19:3650–3663.
Goudet, J. 2001. FSTAT (version 2.9.3): a program to estimate and test gene
diversities and fixation indices. http://www.unil.ch/izea/softwares/fstat.
html. Accessed 10 Feb 2009.
Guo, S. W., and E. A. Thompson. 1992. Performing the exact test of
Hardy-Weinberg proportions for multiple alleles. Biometrics 48:361–372.
Guillot, G., R. Leblois, L. Coulon, and A. C. Frantz. 2009. Statistical
methods in spatial genetics. Molecular Ecology 28:4734–4756.
Hall, L. A., and S. R. Beissinger. 2014. A practical toolbox for design and
analysis of landscape genetics studies. Landscape Ecology 29:1487–1504.
Hanski, I. 2002. Population dynamic consequences of dispersal in local
populations and in metapopulation. Pages 283–298 in J. Clobert, E.
Danchin, A. A. Dhont, and J. D. Nichols, editors. Dispersal. Oxford
University Press, Oxford, United Kingdom.
Hawbacker, T. J., V. C. Radeloff, and M. K. Clayton. 2006. Road
development, housing growth, and landscape fragmentation in northern
Wisconsin: 1937–1999. Ecological Applications 16:1222–1237.
Haynes, G. D., and E. K. Latch. 2012. Identification of novel single
nucleotide polymorphisms (SNPs) in deer (Odocoileus spp.) using the
Bovine SNP50 bead chip. PLoS ONE 7:e36536.
Henderson, D. W., R. J. Warren, J. A. Cromwell, and R. J. Hamilton. 2000.
Response of urban deer to a 50% reduction in local herd density. Wildlife
Society Bulletin 28:902–910.
Hiller, T. L. 2007. Land-use patterns and population characteristics of
white-tailed deer in an agro-forest ecosystem in south central Michigan.
Dissertation, Michigan State University, East Lansing, USA.
Hirth, D. H. 1977. Social behavior of white-tailed deer in relation to habitat.
Wildlife Monographs 53:1–55.
Holderegger, R., and H. H. Wagner. 2008. Landscape genetics. BioScience
58:199–207.
Jenks, G. F. 1967. The data model concept in statistical mapping. Inter
Yearbook Cartography 7:186–190.
Kalinowski, S. T., A. P. Wagner, and M. L. Taper. 2006. ML-Relate: a
computer program for maximum likelihood estimation of relatedness and
relationship. Molecular Ecology Notes 6:576–579.
Kareiva, P., S. Watts, R. McDonald, and T. Boucher. 2007. Domesticating
nature: shaping landscapes and ecosystems for human welfare. Science
316:1866–1869.
Keyghobadi, N., J. Roland, S. F. Matter, and C. Strobek. 2005. Among- and
within-patch components of genetic diversity respond at different rates to
habitat fragmentation: an empirical demonstration. Proceedings of the
Royal Society B 272:553–560.
King, R. T. 1938. The essentials of a wildlife range. Journal of Forestry
36:457–464.
Knowlton, J. L., and C. H. Graham. 2010. Using behavioral landscape
ecology to predict species’ responses to land-use and climate change.
Biological Conservation 143:1342–1354.
Kool, J. T., C. B. Paris, S. Andréfouët, and R. K. Cowen. 2010. Complex
migration and the development of genetic structure in subdivided
populations: an example from Caribbean coral reef ecosystems. Ecography
33:597–606.
Landguth, E. L., S. A. Cushman, M. Murphy, and G. Luikart. 2010.
Relationships between migration rates and landscape resistance assessed using
individual-based simulations. Molecular Ecology Resources 10:854–862.
Landguth, E. L., S. A. Cushman, M. K. Schwartz, K. S. McKelvey, M.
Murphy, and G. Luikart. 2010. Quantifying the lag time to detect barriers
in landscape genetics. Molecular Ecology 19:4179–4191.
Legendre, P., and M. J. Fortin. 2010. Comparison of the Mantel test and
alternative approaches for detecting complex multivariate relationships in
the spatial analysis of genetic data. Molecular Ecology Resources 10:
831–844.
Lichstein, J. W. 2007. Multiple regression on distance matrices: a
multivariate spatial analysis tool. Plant Ecology 118:117–131.
Locher et al.
Deer Spatial Genetic Structure
Long, E. D., D. R. Diefenback, C. S. Rosenberry, B. S. Wallingford, and
M. D. Grund. 2005. Forest cover influences dispersal distance of whitetailed deer. Journal of Mammalogy 86:623–629.
Long, E. D., D. R. Diefenbach, B. D. Wallingford, and C. S. Rosenberry.
2010. Influence of roads, rivers, and mountains on natal dispersal of whitetailed deer. Journal of Wildlife Management 74:1242–1249.
Manel, S., M. S. Schwartz, G. Luikart, and P. Taberlet. 2003. Landscape
genetics: combining landscape ecology and population genetics. Trends in
Ecology and Evolution 18:189–197.
Manel, S., S. Joost, B. K. Epperson, R. Holderegger, A. Storfer, M. S.
Rosenberg, K. T. Scribner, A. Bonin, and M.-J. Fortin. 2010. Perspectives
on the use of geo-environmental data in landscape genetics in the face of
global change. Molecular Ecology 19:3760–3772.
Manel, S., and R. Holderegger. 2013. Ten years of landscape genetics.
Trends in Ecology and Evolution 28:614–621.
Marchinton, R. L., and D. L. Hirth. 1984. Behavior. Pages 129–168 in L. K.
Halls, editor. White-tailed deer: ecology and management. Stackpole
Books, Harrisburg, Pennsylvania, USA.
McRae, B. H., and P. Beier. 2007. Circuit theory predicts gene flow in plant
and animal populations. Proceedings National Academy of Sciences
104:19885–19890.
Michigan Center for Geographic Information. 2002. Michigan Geographic
Framework. Vector digital data. Lansing, Michigan, USA.
Michigan Center for Geographic Information. 2009. Michigan Geographic
Framework: hydrography polygons. Vector digital data. Lansing,
Michigan, USA.
Michigan Department of Natural Resources [MDNR]. 2003. IFMAP/
GAP Upper Peninsula land cover [map]. 1997–2001. 1:64,000. http://
www.dnr.state.mi.us/spatialdatalibrary/sdl2/land_use_cover/2001/IFMAP_
up_landcover.htm. Accessed 4 Nov 2006.
Michigan Department of Natural Resources [MDNR]. 2009. A review of
deer management in Michigan. Michigan Department of Natural
Resources, Lansing, Michigan, USA. http://www.michigan.gov/documents/dnre/WLD_Deer_Mgmt_Plan_Appendix_D-A_Review_of_Deer_
Management_in_Michigan_310657_7.pdf Accessed 15 May 2010.
Michigan Department of Natural Resources [MDNR]. 2010. Michigan
deer management plan. Michigan Department of Natural Resources,
Lansing, Michigan, USA. http://www.michigan.gov/documents/dnre/
WLD_Deer_Management_Plan_FINAL_58_320639_7.pdf. Accessed
12 Nov 2014.
Milligan, B. G. 2003. Maximum-likelihood estimation of relatedness.
Genetics 163:1153–1167.
Minor, E. S., and D. L. Urban. 2007. A graph-theory framework for
evaluating landscape connectivity and conservation planning. Conservation Biology 22:297–307.
Miranda, B. F., and W. F. Porter. 2003. Statewide habitat assessment for
white-tailed deer in Arkansas using satellite imagery. Wildlife Society
Bulletin 31:715–726.
Morris, D. W., J. E. Diffendorfer, and P. Lundberg. 2004. Dispersal among
habitats varying in fitness: reciprocating migration through ideal habitat
selection. Oikos 107:559–579.
Morrison, M. L., B. Marcot, and W. Mannan. 2006. Wildlife-habitat
relationships: concepts and applications. Island Press, Washington, D.C.,
USA.
Murphy, M. A., R. Dezzani, D. S. Pilliod, and A. Storfer. 2010. Landscape
genetics of high mountain frog metapopulations. Molecular Ecology
19:3634–3649.
Nixon, C. M., M. W. McClain, and K. R. Russell. 1970. Deer food habits
and range characteristics in Ohio. Journal of Wildlife Management
34:870–886.
Peakall, R., and P. E. Smouse. 2012. GenAlEx 6.5. Bioinformatics 28:2537–2539.
Pérez-Espona, S., F. J. Pérez-Barberı́a, and J. E. Mcleod. 2008. Landscape
features affect gene flow of Scottish highland red deer (Cervus elaphus).
Molecular Ecology 17:981–996.
Purrenhage, J., P. H. Niewiarowski, and F. B. Moore. 2009. Population
structure of spotted salamanders (Ambystoma maculatum) in a fragmented
landscape. Molecular Ecology 18:235–247.
Pusateri, J. S. 2003. White-tailed deer population characteristics and
landscape use patterns in southwestern Lower Michigan. Thesis,
Michigan State University, East Lansing, USA.
Raymond, M., and F. Rousset. 1995. An exact test for population
differentiation. Evolution 43:1243–1286.
193
Rice, W. R. 1989. Analyzing tables of statistical tests. Evolution 43:
223–225.
Robinson, S. J., M. D. Samuel, D. L. Lopez, and P. Shelton. 2012. The walk
is never random: subtle landscape effects shape gene flow in a continuous
white-tailed deer population in the Midwestern United States. Molecular
Ecology 21:4190–4205.
Ronce, O., I. Olivieri, and J. Clobert. 2001. Perspectives on the study of
dispersal evolution. Pages 341–357 in J. Clobert, E. Danchin, A. Dhondt,
and J. D. Nichols, editors. Dispersal. Oxford University Press, Oxford,
United Kingdom.
Roseberry, J. L., and A. Woolf. 1998. Habitat-population density
relationships for white-tailed deer in Illinois. Wildlife Society Bulletin
26:252–258.
Rosenberg, M. S., and C. D. Anderson. 2011. PASSaGE: pattern analysis,
spatial statistics and geographic exegesis. Version 2. Methods in Ecology
and Evolution 2:229–232.
Rousset, F. 1997. Genetic differentiation and estimation of gene flow from
F-statistics under isolation by distance. Genetics 145:1219–1228.
Rousset, F. 2008. GENEPOP 2007: a complete reimplementation of the
GENEPOP software for Windows and Linux. Molecular Ecology
Resources 8:103–106.
Scribner, K. T., M. S. Petersen, R. L. Fields, J. Pearce, and R. K. Chesser.
2001. Sex-biased gene flow in spectacled eiders (Anatidae): inferences
from molecular markers with contrasting modes of inheritance. Evolution
55:2105–2115.
Shi, H., E. J. Laurent, J. LeBouton, L. Racevskis, K. R. Hall, M. Donovan,
R. V. Doepke, M. B. Walters, F. Lupi, and J. Liu. 2006. Local spatial
modeling of white-tailed deer distribution. Ecological Modeling 190:
171–189.
Sommers, L. M. 1977. Atlas of Michigan. Michigan State University Press,
East Lansing, Michigan, USA.
Spear, S. F., N. Balkenhol, M.-J. Fortin, B. McRae, and K. T. Scribner.
2010. Use of resistance surfaces for landscape genetic studies: considerations for parameterization and analysis. Molecular Ecology 19:
3576–3591.
194
Storfer, A., M. A. Murphy, J. S. Evans, C. S. Goldberg, S. Robinson,
S. F. Spear, R. Dezzani, E. Delmelle, L. Vierling, and L. P. Waits.
2007. Putting the ‘landscape’ in landscape genetics. Heredity
98:128–142.
Trombulak, S. C., and C. A. Frissell. 2000. Review of ecological effects of
roads on terrestrial and aquatic communities. Conservation Biology
14:18–30.
Underhill, J. E., and P. G. Angold. 2000. Effects of roads on wildlife in an
intensively modified landscape. Environmental Review 8:21–39.
Urban, D., E. S. Minor, and E. A. Treml. 2009. Graph models of habitat
mosaics. Ecological Letters 12:260–273.
Verme, L. J. 1973. Movements of white-tailed deer in Upper Michigan.
Journal of Wildlife Management 37:545–552.
Wagner, H. H., and M.-J. Fortin. 2005. Spatial analysis of landscapes:
concepts and statistics. Ecology 86:1975–1987.
Wagner, A., S. Creel, and S. T. Kalinowski. 2006. Estimating relatedness
and relationships using microsatellite loci with null alleles. Heredity
97:336–345.
Waples, R. S., and O. Gaggiotti. 2006. What is a population? An empirical
evaluation of some methods for identifying the number of gene pools and
their degree of connectivity. Molecular Ecology 15:1419–1439.
Weir, B. S., and C. C. Cockerham. 1984. Estimating F-statistics for the
analysis of population structure. Evolution 38:1358–1370.
White, G. C., and R. A. Garrot. 1990. Analysis of wildlife radiotracking
data. Academic Press, San Diego, California, USA.
Wilson, G. A., C. Strobeck, and L. Wu. 1997. Characterization of
microsatellite loci in caribou, Rangifer tarandus, and their use in other
artiodactyls. Molecular Ecology 6:697–699.
With, K. A., R. H. Gardner, and M. G. Turner. 1997. Landscape
connectivity and population distributions in heterogeneous environments.
Oikos 78:151–169.
Wright, S. 1943. Isolation by distance. Genetics 28:114–138.
Associate Editor: Emily Latch.
The Journal of Wildlife Management
79(2)